Dynamic Difficulty Adjustment of Game AI by MCTS for the game Pac-Man

Abstract

Dynamic Difficulty Adjustment (DDA) of Game AI aims at creating a satisfactory game experience by dynamically adjusting intelligence of game opponents. It can provide a level of challenge that is tailored to the player's personal ability. The Monte-Carlo Tree Search (MCTS) algorithm can be applied to generate intelligence of non-player characters (NPCs) in video games. And the performance of the NPCs controlled by MCTS can be adjusted by modulating the simulation time of MCTS. Hence the approach of DDA based on MCTS is proposed based on the application of MCTS. In this paper, the prey and predator game genre of Pac-Man is used as a test-bed, the process of creating DDA based on MCTS is demonstrated and the feasibility of this approach is validated. Furthermore, to increase the computational efficiency, an alternative approach of creating DDA based on knowledge from MCTS is also proposed and discussed.